Optimizing Urban Green Roofs: An Integrated Framework for Suitability, Economic Viability, and Microclimate Regulation

Abstract

Urban areas face significant challenges from heat islands, stormwater, and air pollution, yet green roof adoption is hindered by feasibility and economic uncertainties. This study proposes an integrated framework to optimize green roof strategies for urban sustainability. We combine deep learning for rooftop suitability screening, comprehensive ecosystem service valuation, life-cycle cost–benefit analysis under varying policy scenarios, and ENVI-met microclimate simulations across Local Climate Zones (LCZ). Using Dalian’s core urban districts as a case study, our findings reveal that all three green roof types (extensive, semi-intensive, and intensive) are economically viable when policy incentives and ecological values are fully internalized. Under the ideal scenario, intensive roofs yielded the highest long-term returns with a payback period of 4 years, while semi-intensive roofs achieved the greatest cost-effectiveness (BCR = 4.57) and the shortest payback period of 3 years; extensive roofs also reached break-even within 4 years. In contrast, under the realistic market-only scenario, only intensive roofs approached break-even with an extended payback period of 23 years, whereas extensive and semi-intensive systems remained unprofitable. Cooling performance varies by LCZ and roof type, emphasizing the critical role of urban morphology. This transferable framework provides robust data-driven decision support for green infrastructure planning and targeted policymaking in high-density urban environments.

1. Introduction

With the acceleration of urbanization and the intensification of climate change, environmental issues such as the urban heat island (UHI) effect, air pollution, stormwater runoff pressure, and biodiversity loss have become increasingly severe, posing serious threats to public health and urban sustainability [1,2,3]. Against this backdrop, urban green infrastructure has attracted growing attention as an effective means of enhancing ecosystem services and improving urban environmental quality [4]. Among various forms of green infrastructure, green roofs—integrating multiple functions such as energy conservation, temperature reduction, stormwater management, and air purification—are not only regarded as an effective strategy for improving building sustainability but also provide ecological, economic, and social benefits at the urban scale [5,6,7]. In recent years, green roofs have played an increasingly important role in urban planning and practice worldwide, becoming a key measure in promoting sustainable urban development [8].

Despite the widely recognized ecological and social benefits of green roofs, their large-scale implementation in cities, including those in China, still faces multiple barriers [9]. For instance, building characteristics such as type, age, structural design, roof slope, and load-bearing capacity, along with environmental conditions such as sunlight exposure and ventilation, significantly affect both the feasibility of retrofitting and the effectiveness of greening efforts [10,11]. Scientifically identifying and accurately selecting suitable rooftops is essential to maximizing benefits and optimizing resource allocation in green roof projects. To date, rooftop suitability assessments have primarily focused on physical indicators, while often overlooking practical considerations such as long-term maintenance feasibility and rooftop microclimatic conditions. Shao et al. assessed rooftop suitability based on physical indicators such as roof slope and structural capacity [12], Velázquez et al. further prioritized areas based on environmental pollution and human activity [13]. Tomás et al. employed computer-vision techniques and geospatial analysis, integrating aerial imagery within MATLAB and QGIS workflows to delineate rooftop areas suitable for green roof installation [14]. Likewise, Francisco et al. combined remote sensing and LiDAR data to map green roof deployment potential in Granada, Spain, using roof slope and rooftop area as the principal screening criteria [15]. Although these studies have certain limitations—particularly in addressing management feasibility and microclimatic conditions—they nonetheless provide valuable methodological references for the siting and prioritization of urban green roofs. Building upon these foundations, this study integrates a broader range of criteria to develop a more comprehensive and practical evaluation framework for rooftop suitability in complex urban environments.

In academic literature, green roofs are commonly classified into three types based on substrate depth, plant types, and maintenance requirements: extensive [16], semi-intensive [17], and intensive systems [17,18]. Although these three types differ markedly in terms of construction conditions, ecological contributions, and maintenance complexity, many previous studies focused on a single roof type or did not clearly differentiate the ecological and economic characteristics of each system. This has limited the potential for optimizing green roof configurations across varying urban contexts.

Moreover, while green roofs are widely recognized as important tools for improving urban thermal environments and enhancing ecosystem services, large-scale adoption remains constrained by high initial investment costs and uncertain economic returns [19]. Current evaluation systems still have significant limitations. On one hand, most studies emphasize individual ecological indicators—such as carbon sequestration [14,20,21], stormwater management [22,23], or air purification [24,25,26,27]—without comprehensively integrating the diverse ecological and economic benefits of green roofs. As a result, assessments of return on investment and market potential often lack completeness and practical relevance. On the other hand, as urban spatial structures grow more complex, the ecological and thermal benefits of green roofs display considerable heterogeneity across different spatial contexts [28]. However, building on prior work that established LCZ as a useful organizing framework, most integrated studies have combined a limited subset of methods—for example, LCZ-based microclimate simulations to evaluate green roofs [29,30], or cost–benefit analysis [31]. While these contributions demonstrate the value of LCZ-resolved assessment, they rarely deliver an end-to-end pipeline that starts from a deployable rooftop inventory and carries through to LCZ-resolved microclimate performance and life-cycle economics under policy scenarios. Extending this line of work, our study raises the level of integration by linking rooftop suitability screening, LCZ-resolved ENVI-met simulations, and dual-scenario life-cycle CBA within a single, reproducible workflow, organized by the three roof types (extensive, semi-intensive, and intensive) to produce actionable configuration and policy guidance.

To address three unresolved gaps—(1) the lack of a deployable rooftop-suitability baseline that goes beyond physical indices to include accessibility and on-roof microclimate; (2) the absence of policy-sensitive, life-cycle economic evaluation that integrates multiple ecosystem services; and (3) the scarcity of a system-level framework that integrates rooftop suitability screening, LCZ-resolved ENVI-met simulations, and dual-scenario life-cycle cost–benefit analysis—we use Dalian’s Zhongshan and Xigang districts as case areas and develop an integrated framework linking these components end-to-end.

Our core question is: within heterogeneous urban forms (captured by LCZs), how should green-roof types and their spatial placement be selected—based on rooftop suitability—to simultaneously maximize ecological benefits and ensure economic feasibility?

To tackle this central question, the study employs a three-pronged approach:

• Based on a defined indicator system, identify urban spaces suitable for green roof development by integrating multi-source spatial data, deep learning-based object detection, and remote sensing.
• For the three green roof strategies, develop two investment evaluation models—an “ideal scenario” (a comprehensive environmental-subsidy framework is in place: the ecological value of green roofs is monetized (internalized) and disbursed as subsidies to residents, construction firms/developers, or local governments) and a “realistic scenario” (environmental subsidy mechanisms are limited: only energy-saving and demand-reduction benefits translate into monetary returns, while broader ecosystem services are not compensated)—to conduct life-cycle cost–benefit analyses and assess their advantages and disadvantages from different investment perspectives.
• Apply the LCZ classification method in combination with the ENVI-met microclimate simulation tool to systematically compare the thermal regulation effects of different greening strategies under representative LCZ types.

Through multi-type and multi-scenario benefit evaluation and microclimate simulation, this study revealed differences in ecological contributions and economic returns among various green roof strategies and proposed configuration recommendations based on spatial prioritization and type optimization, within a framework designed to be adaptable to other cities with different climates, morphology, and policy environments. The research not only enriched the theoretical framework for spatial evaluation of green roofs but also provided data-driven decision support for the scientific planning of green infrastructure in high-density urban areas.

2. Materials and Methods

The detailed research process is illustrated in Figure 1. First, based on key evaluation indicators, buildings suitable for green roof retrofitting within the study area were systematically identified. Next, using the estimated suitable roof area, the benefit transfer method was applied to integrate literature-based parameters and quantify the ecological functions and economic value of the three green roof types (extensive, semi-intensive, and intensive). A life-cycle cost–benefit analysis was then conducted under both ideal and realistic scenarios to assess investment feasibility. Finally, representative sample areas were selected based on LCZ classification, and the ENVI-met model was used to simulate pedestrian-level cooling effects of different roof types across various urban morphologies.

2.1. Study Area

This study selected Zhongshan and Xigang Districts in Dalian, China, as representative high-density urban core areas for evaluation (Figure 2). Both districts are located in the city center and serve as major hubs for economic, administrative, and cultural functions in Dalian [32].

Over the past four decades, rapid urbanization in Dalian has led to high land-use intensity and a large proportion of impervious surfaces in urban areas, significantly contributing to the urban heat island effect [33]. Based on ENVI 6.0 processing of Landsat 8 satellite data (Figure 2) and supporting literature, the dense high-rise zones in Zhongshan District and the port-adjacent industrial zones in Xigang District exhibited elevated surface temperatures, making them among the most affected areas by urban thermal environment issues in Dalian [32].

Additionally, this region held substantial potential for green roof implementation. On one hand, the high density of mid- and high-rise buildings provided a solid spatial foundation for rooftop greening. On the other hand, the presence of historically protected buildings and lightweight industrial structures introduced constraints representative of common challenges in green roof retrofitting. Therefore, selecting this region as the study area offered strong empirical relevance and practical reference value for broader applications.

2.2. Rooftop Suitability for Green Roof Construction

The three green roof types—intensive, semi-intensive, and extensive—differ significantly in substrate depth, vegetation type, and load-bearing weight, thereby imposing distinct structural and environmental requirements on buildings [34,35]. To scientifically and efficiently identify buildings suitable for green roof retrofitting, this study established a roof suitability assessment system based on multiple constraint factors. The six key evaluation factors—building structure, rooftop accessibility, rooftop slope, building historical value, roof sunlight exposure, and roof surface wind speed—were selected based on national building codes, green roof design guidelines, and previous roof suitability assessments [36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59]. These factors collectively account for structural safety, legal feasibility, and the environmental conditions critical to the long-term performance of green roofs.

All factors were treated as hard constraints, meaning that if a building failed to meet the minimum threshold for any single element, it was deemed unsuitable for green roof development, thereby minimizing the risk of overestimating the implementation potential. The specific thresholds for each criterion are presented in

Table 1. Suitability criteria and indicator settings for green roof implementation.

Indicator	Evaluation Criteria	Method	References
Building Structure	Non-industrial (excluding lightweight steel structures)	Land-use and remote sensing	[36,37,38,39]
Roof Accessibility	Roof accessibility	Remote sensing and object detection	[40,41,42,43]
Roof Slope	Flat roof (slope ≤ 20°)	Remote sensing and object detection	[44,45,46]
Building Historical Value	Not a designated historic building	Government open data	[47,48,49,50,51,52]
Roof Sunlight Exposure	Daily sunlight ≥ 4 h on summer solstice	Ladybug	[53,54,55,56]
Roof Surface Wind Speed	Mean wind speed < 20 m/s	Phoenics	[57,58,59]

, and the detailed filtering methods are explained in Section 2.2.1, Section 2.2.2, Section 2.2.3, Section 2.2.4, Section 2.2.5 and Section 2.2.6. Based on this framework, all eligible rooftops in the study area were systematically identified, and their spatial distribution and total area were quantified using QGIS (Version 3.40.1 Release data 22 November 2024) software.

2.2.1. Evaluation Elements 1: Building Structure

Building structure determined whether a roof had sufficient load-bearing capacity to support a green roof system [36]. Different green roof types imposed varying structural load requirements on rooftops (

Table 2. Comparison of three types of green roofs (adapted from [37]).

Type of Green Roof	Vegetation Type	Substrate Depth	Weight
Extensive	Moss-Grass	60–200 mm	60–150 kg/m2
Semi-intensive	Grass-Shrub	120–500 mm	120–200 kg/m2
Intensive	Shrub-Tree	150–750 mm	180–500 kg/m2

) [37]. Generally, buildings with reinforced concrete structures whose main beams had a load-bearing capacity over 2.98 kN/m², and concrete slab structures estimated to support 8–10 kN/m², met the structural requirements for extensive to intensive green roofs [38].

Due to the absence of detailed building load records in the study area, this research used structural type as a proxy. In Dalian, most modern urban buildings typically met these structural load requirements, while industrial factory buildings, often constructed with lightweight steel structures, had relatively low load-bearing capacities (0.3–0.5 kN/m²) and were unsuitable for green roof installation [39]. Thus, this study considered non-industrial buildings to satisfy the basic condition for structural suitability.

2.2.2. Evaluation Elements 2: Rooftop Accessibility

Roof accessibility was a key condition for the long-term maintenance and management of green roofs [40,41]. During the early stages of construction, maintenance tasks such as irrigation system checks, pest and disease control, and substrate replenishment were required [42]. Accessible roofs typically had adequate structural capacity and entry points, making them more suitable for regular manual operations. According to regulations from the Ministry of Housing and Urban–Rural Development of the People’s Republic of China, accessible roofs must have a minimum load-bearing capacity of 2.0 kN/m² [39], which also served as an indirect indicator of structural adequacy.

To enable large-scale, automated identification of roof accessibility, this study developed a deep learning-based object detection model. The training dataset consisted of manually labeled rooftop images collected via Google Earth. A total of 500 remote sensing images were annotated using the Labelme tool. Each rooftop was assigned one of the following three labels: “accessible roof”, “inaccessible roof”, or “sloped roof” [43]. The dataset was divided into training, validation, and test sets at a ratio of 8:1:1. The model was trained using the YOLOv5 architecture for object detection. The input image size was set to 640 × 640, with 100 training epochs and a batch size of 8. The initial learning rate was 0.01 and dynamically adjusted during training to accelerate convergence and reduce oscillation. The momentum parameter was set to 0.937 to enhance optimization efficiency and gradient stability. The Intersection over Union (IoU) threshold for positive-negative sample matching was 0.7. The optimizer type was set to auto-selection, allowing adaptive configuration of the optimal training strategy. Model performance was evaluated using mean Average Precision (mAP@0.5, IoU = 0.5). The YOLOv5 model achieved mAP@0.5 = 88.83% on the test set, after which predictions were manually checked before downstream suitability analysis.

To situate this performance within the current state of the art for urban rooftop recognition from aerial imagery, we compare against three recent studies. First, Boccalatte et al. [44] quantified urban solar-potential losses by detecting rooftop superstructures over ~35,000 buildings: a U-Net (ResNet-152) segmentation model reached IoU = 0.40 for obstruction masks, and downstream “free/obstructed” roof-segment classification achieved 85–91% accuracy after cross-validation. Although their primary metric (IoU for masks) differs from our detection mAP, their reported classification-level accuracy in the mid-to-high 80% range provides a relevant benchmark for rooftop-object recognition at city scale.

Second, Chen et al. [45] proposed a GAN-enhanced pipeline (colorization + super-resolution) for historical black-and-white aerial photos and reported mAP > 85% for rooftop detection with YOLO-family models, demonstrating that well-configured detectors consistently reach mid-80% or higher on rooftop tasks despite challenging imagery.

Third, Mei et al. [46] addressed roof-edge polygonization with a YOLOv8-OBB–based edge detector, reporting raster-level mIoU ≈ 0.85–1.00 on SGA/Melville datasets and ovIoU ≈ 0.97 for vectorized roofs—evidence that modern detectors attain high agreement for rooftop structures under high-resolution aerial data. While edge-polygon metrics are not identical to object-detection mAP, they reinforce that accuracies in the mid-80% and above are typical for current rooftop recognition pipelines, which is consistent with our results.

Taken together, these external benchmarks show that our rooftop-accessibility detector operates within the prevailing accuracy range for city-scale rooftop recognition—and in several cases at the upper end—thus providing a reliable basis for subsequent suitability screening and policy analysis.

2.2.3. Evaluation Elements 3: Rooftop Slope

Roof slope affects the stability of green roof systems and the complexity of their construction, with steeper slopes generally increasing construction difficulty [41]. High-slope roofs are prone to issues such as substrate slippage, water loss, and poor plant establishment [47,48]. These roofs often required additional anchoring systems or structural modifications, significantly increasing both construction and maintenance costs [39,47]. Therefore, ecological performance assessments for green roof retrofitting commonly recommend prioritizing flat roofs with slopes in the 0–20° range [49]. Similarly, building regulations issued by the Ministry of Housing and Urban–Rural Development of the People’s Republic of China explicitly advised prioritizing flat roofs for green roof implementation [39].

The same 500 high-resolution remote sensing images described in Section 2.2.2 were used here. Among the annotated rooftops, those labeled as “sloped roof” were used to identify non-flat structures, while flat roofs included both “accessible” and “inaccessible” types. The YOLOv5 model was trained on this unified three-class dataset, enabling simultaneous detection of roof accessibility and slope.

2.2.4. Evaluation Elements 4: Building Historical Value

Green roof construction typically involves additional structural loading, waterproofing modifications, and root barrier installation to prevent vegetation root intrusion [50]. These procedures may pose risks to the materials and structural integrity of historically protected buildings [51]. According to relevant Chinese laws and regulations, green roof retrofitting on buildings with historical value may violate principles such as “no alteration to the original state of cultural relics” and “minimum intervention”, leading to compliance risks [52,53,54]. Therefore, this study included “whether a building is historically protected” as a restrictive criterion in the green roof suitability assessment. The scope of historically protected buildings within the study area was identified using open-access data released by the Dalian municipal government.

2.2.5. Evaluation Elements 5: Roof Sunlight Exposure

Plant growth is highly sensitive to light availability. In densely built urban environments, shading from nearby high-rise buildings may result in inadequate rooftop sunlight, affecting the stability and aesthetic performance of green roof systems [10,12,55]. In this study, Roof Sunlight Exposure was simulated using the Ladybug Tools plugin within the Rhino-Grasshopper platform [56]. First, building data with elevation attributes was imported into Rhino to construct accurate 3D building models. Subsequently, based on typical meteorological year (TMY) climate data for Dalian, rooftop sunlight exposure was simulated for the summer solstice (June 21) [57]. Referring to studies on local herbaceous plant growth under different light conditions [58], a restrictive threshold of “no less than 4 h of sunlight on the summer solstice” was adopted. Rooftops not meeting this threshold were considered unsuitable for green roof development.

2.2.6. Evaluation Elements 6: Roof Surface Wind Speed

Wind speed significantly affects plant morphology and physiological function. Excessive rooftop wind can increase water loss, reduce photosynthetic efficiency, and lead to soil moisture depletion and changes in nutrient availability [59,60]. When wind speeds exceed 20 m/s, mechanical damage to plant tissues may occur [61]. Therefore, this study used whether Roof Surface Wind Speed reached or exceeded 20 m/s as a restrictive criterion for green roof suitability.

The rooftop wind environment was simulated using Phoenics (Version 2019 v1.0,Release data March 2019), a Computational Fluid Dynamics (CFD) software developed by CHAM [62]. The 3D building models created in Section 2.2.5 were imported into the software, along with TMY climate data. Dominant wind directions and boundary conditions were set to simulate wind flow through dense urban clusters. Based on the simulation output, buildings with Roof Surface Wind Speeds ≥ 20 m/s were excluded from suitability.

2.3. Estimation of Ecological and Economic Benefits

2.3.1. Per Unit Area Benefit Parameters and Total Benefit Calculation

Given the absence of empirical ecological data specific to green roofs in Dalian, this study drew on a set of widely cited, methodologically rigorous domestic and international studies [63,64,65,66,67,68,69] to enhance the scientific validity and comparability of benefit estimation. Per-unit-area benefit parameters were compiled for five key ecological functions across three green-roof typologies—extensive, semi-intensive, and intensive (

Table A1. Unit area benefit values of five ecological functions of green roofs.

Types of Ecosystem Services	Green Roof Type	Per-Unit-Area Ecological Benefit Values	Source
Carbon Sequestration	Extensive	2.21 kgCO2/m2/y	[63]
	Semi-intensive	5.03 kgCO2/m2/y	[64]
	Intensive	6.58 kgCO2/m2/y	[64]
Oxygen Release	Extensive	1.61 kgO2/m2/y	[63]
	Semi-intensive	5.34 kgO2/m2/y	[66]
	Intensive	7.19 kgO2/m2/y	[63]
Air Pollution Reduction	Extensive	23.2 µg/m3/y	[65]
	Semi-intensive	34.82 µg/m3/y	[65]
	Intensive	37.1 µg/m3/y	[65]
Stormwater Retention	Extensive	45%	[67]
	Semi-intensive	85%	[68]
	Intensive	85%	[68]
Building Energy Conservation	Extensive	6.17 kWh/m2/y	[69]
	Semi-intensive	17.45 kWh/m2/y	[69]
	Intensive	21.5 kWh/m2/y	[69]

).

Specifically, carbon sequestration and oxygen release coefficients were drawn from studies conducted in Wuhan, Shenyang, and northern urban residential contexts; PM2.5 adsorption rates were based on field measurements in Xuzhou; stormwater retention ratios were obtained from international reviews of green roof hydrological performance; and energy-saving estimates were adapted from modeling studies in Mediterranean climate zones.

Although some of the source cities (e.g., Wuhan and Xuzhou) differ from Dalian in climatic classification, the selected parameters are considered broadly applicable based on two main considerations. First, all values were conservatively estimated using minimum or average figures, in order to avoid overestimating ecosystem performance—an approach consistent with common practices in green infrastructure modeling. Second, as green roofs are engineered ecological systems, their functional outputs are primarily influenced by plant species, substrate depth, and maintenance practices, rather than by macroclimatic conditions. This makes the cross-regional transfer of parameters methodologically sound.

Overall, these parameters are considered methodologically transparent, transferable, and sufficiently adaptable for simulating the ecological benefits of green roofs under the temperate monsoon climate of Dalian. It should be noted that these values are intended to provide reasonably accurate estimates for comparative analysis, rather than precise measurements of absolute ecological outcomes.

To evaluate the ecological value of green roofs in the study area, the total suitable rooftop area (as determined in Section 3.1) was multiplied by the per-unit-area benefit parameters for each green roof type and each ecosystem function listed in Table A1. This yielded total ecological benefit estimates for each green roof category.

2.3.2. Monetary Conversion

To monetize these ecological benefits, this study adopted the valuation approach proposed by Teotónio et al. [70] in their assessment of green roofs in Lisbon, Portugal. The five ecological services—Carbon Sequestration, Oxygen Release, Air Pollution Reduction, Stormwater Retention, and Building Energy Conservation—were each converted into economic values, forming a localized ecosystem service valuation system adapted to Dalian’s market context. The monetary conversion parameters were derived from the latest local market prices. Among these, annual rainfall captured by suitable rooftops was estimated using the 2024 average annual precipitation for Dalian and used to calculate the economic value of stormwater retention.

It is important to note that among these services, only Building Energy Conservation—measured through reduced electricity consumption—could directly yield financial benefits for building owners via energy cost savings [71]. The remaining services are considered public goods [72] and are not currently incorporated into a mature market compensation system. Their value realization depends on public funding or supportive incentive policies [73].(

Table A2. Recent price data related to ecosystem services in Dalian.

Types of Ecosystem Services	Price	Source
Carbon Sequestration	0.017 $/kgCO2	[74]
Oxygen	0.055 $/kg	Quotations from local artificial oxygen manufacturers
Air Pollutant Treatment	0.608 $/µg	[75]
Water	0.402 $/m3	[76]
Electrical Power	0.315 $/kWh	[77]

, [74,75,76,77])

2.3.3. Construction and Maintenance Cost Assumptions

Green roof costs primarily include initial construction investments and ongoing maintenance expenses, both of which are influenced by local market conditions, labor pricing, and project types. However, due to the lack of publicly available and systematic green roof cost datasets in Dalian, this study adopted cost parameters published by the Beijing Municipal Bureau of Landscaping (

Table A3. Construction and maintenance costs (adapted from [78]).

	Extensive	Semi-Intensive	Intensive
Standard waterproofing layer ($/m2)	9.724	8.335	9.724
Root-resistant waterproofing layer ($/m2)	14.586	12.502	14.586
Protective layer ($/m2)	2.083	1.666	2.083
Drainage and retention layer ($/m2)	3.472	3.472	4.167
Filter layer ($/m2)	1.389	1.666	1.944
Planting substrate ($/m2)	6.945	34.728	34.728
Plant materials and installation ($/m2)	8.334	13.335	19.726
Irrigation system ($/m2)	2.083	1.666	2.083
Total construction costs ($/m2)	48.62	70.43	89.045
Maintenance costs ($/m2/y)	2.014	2.917	2.917

As of 17 August 2025, the exchange rate was 1 USD = 7.1986 CNY.

) [78].

Although the data originate from Beijing, their application to the Dalian context is well justified. Both cities are located in northern coastal China and fall within the same Warm Temperate Monsoon Climate Zone. In addition, they share comparable urbanization levels, building typologies, and development intensity. The construction industry structure, building materials supply chains, and labor pricing mechanisms in these tier-one and tier-two northern cities are regionally aligned. As such, cost benchmarks formulated by Beijing authorities are commonly referenced in policy formulation and project planning across northern China and are considered a credible proxy in regions like Dalian, where local equivalents are unavailable.

Specifically, cost data for basic green roofs in Beijing were used as parameters for extensive green roofs in this study. For semi-intensive and intensive green roofs, the minimum and maximum construction and maintenance costs for garden-style green roofs were selected to represent their respective cost parameters. A 40-year life cycle was assumed to enhance the scientific rigor of the economic evaluation, with annual maintenance expenses included. Structural reconstruction and other exceptional cases were not considered.

2.3.4. Cost–Benefit Analysis Method and Scenario Setting

Cost–Benefit Analysis (CBA) is a decision-making tool used to determine whether a project, policy, or investment is economically justified. Its core principle is to compare total benefits and costs to evaluate economic feasibility [79,80]. Two commonly used indicators in CBA are Net Present Value (NPV) and Benefit–Cost Ratio (BCR).

NPV represents the difference between the present value of cash inflows and outflows over time and is widely used in investment planning to assess projected profitability [81,82]. Its formula is shown in Equation (1). The Discount Rate, a key parameter, reflects the present value of future funds [83]. Following Feng et al. [84], this study adopted a uniform discount rate of 3% for all three green roof types.

$$\mathrm{N}\mathrm{P}\mathrm{V} = \sum _{t=0}^T{\displaystyle \frac{(B_t-C_t)}{{(1+r)}^t}}$$

(1)

where NPV = net present value, B_t = benefit in year t, C_t = cost in year t, r = discount rate, t = time period (0 to T), and T = total years.

BCR is a relative indicator that compares the present value of a project’s benefit stream to its cost stream [85,86]. The formula is provided in Equation (2):

$$\mathrm{B}\mathrm{C}\mathrm{R} = {\displaystyle \frac{\sum _{\mathrm{t}=0}^{\mathrm{T}}\frac{{\mathrm{B}}_{\mathrm{t}}}{{(1+\mathrm{r})}^{\mathrm{t}}}}{\sum _{\mathrm{t}=0}^{\mathrm{T}}\frac{{\mathrm{C}}_{\mathrm{t}}}{{(1+\mathrm{r})}^{\mathrm{t}}}}}$$

(2)

where NPV = net present value, B_t = benefit in year t, C_t = cost in year t, r = discount rate, t = time period (0 to T), and T = total years.

In this study, the five ecological values listed in Table A1were integrated into a unified economic benefit system, and both NPV and BCR were calculated. Among these, the Building Energy Conservation benefit—measured by energy cost savings—could be directly translated into financial returns for building owners [87]. In contrast, the remaining ecological benefits are typical public goods and, under current market mechanisms, do not yield direct revenue [88]. Consequently, energy savings often serve as the primary economic driver for green roof investment, while the realization of other ecological values depends heavily on public support and fiscal incentives.

To comprehensively assess economic feasibility, two cost–benefit estimation scenarios were established:

• Ideal Scenario: All five ecological benefits were monetized, reflecting the full societal return of green roof systems.
• Realistic Scenario: Only energy cost savings from Building Energy Conservation were considered, simulating expected returns without policy subsidies or ecosystem payment schemes.

These scenarios enabled a systematic evaluation of green roof investment viability under different policy and market conditions, offering multi-dimensional references for both government planning and private-sector investment.

2.4. LCZ Classification and Selection of Sample Areas

Another essential value of green roofs lies in their capacity to regulate the local thermal environment [89], a function that has gained prominence amid growing urban heat risks, especially in terms of pedestrian comfort and public health [90]. Accordingly, this study established a second evaluation dimension: simulating the cooling performance of different green roof strategies within typical urban structural zones, with a focus on localized heat exposure mitigation.

While the Building Energy Conservation analysis (Section 2.3.2) focused on reducing indoor temperatures, this section introduced ENVI-met microclimate simulations to assess ambient temperature changes at pedestrian height. Although differing in scale and scope, both assessments contributed to understanding the total value of green roof systems: the former indicated direct economic benefits, and the latter captured ecological services within public space. Together, they supported integrated optimization of planting strategies.

To more accurately evaluate thermal regulatory performance across varied urban zones, the LCZ classification system was used. LCZ is an urban thermal zoning framework that accounts for land cover and human activity characteristics, capturing spatial heterogeneity in building height, density, and surface materials. It is widely adopted in urban climate studies [23].

Compared with other urban morphological frameworks, the LCZ classification is more aligned with the objectives of this study. While systems such as UST emphasize universality and structural generalization [24]. As this study focuses on assessing the ecological and thermal benefits of green roofs, LCZ provides both cross-city comparability and climate-relevant fine-scale differentiation, making it the more suitable choice.

Following the methodology of Stewart and Oke [23], this study utilized the latest remote sensing imagery, GIS building height data, and GIS land-use data, in conjunction with LCZ data for Dalian from the WUDAPT portal [24], to generate an LCZ map for the city. Through the integration and analysis of these multiple data sources, nine distinct urban structural types were identified (Figure 3), resulting in a more precise and up-to-date LCZ classification.

Among these, LCZ1 (compact high-rise), LCZ2, and LCZ8 were selected as representative zones for subsequent ENVI-met simulations. This selection is based on the large coverage of these three LCZ categories within the Zhongshan and Xigang districts of Dalian, as well as their significant concentration of suitable rooftops, which are crucial for urban heat risk analysis. Furthermore, according to studies on heat exposure in Dalian, LCZ1, LCZ2, LCZ8, and LCZ10 (industrial zone) exhibit high levels of heat risk [25]. It is worth noting that although LCZ10 shows the highest heat risk, it was excluded from this study due to its industrial characteristics (building structures and land-use properties) that are not suitable for retrofitting with green roofs.

2.5. ENVI-Met Microclimate Simulation

Based on the representative LCZ identified in Section 2.4, this study employed the ENVI-met model to simulate ambient temperatures at pedestrian height (0.8 m) and evaluate the thermal regulation performance of different green roof types across various LCZ categories. ENVI-met is a three-dimensional, non-hydrostatic microclimate model grounded in physical principles. It captured the spatial heterogeneity of urban structures in detail and accurately simulated the influence of vegetation types and physiological processes on microclimate conditions [91].

The model also supported the integration of measured meteorological data, enabling the use of realistic boundary conditions for urban heat island analysis [92,93]. Compared to other thermal environment simulation models such as Weather Research and Forecasting (WRF), ENVI-met is more suitable for simulating the thermal environment at small to medium urban scales, particularly at the neighborhood or street-block level. Therefore, ENVI-met was well suited to the objectives of this study and served as the appropriate simulation tool for analyzing the thermal regulatory effects of green roofs.

We used ENVI-met (Version 5.6.1, Release date May 2024), a version that has been independently validated for ur ban greening applications—including green roofs—in climates comparable to our study area, showing good agreement for near-surface air temperature, and mean radiant temperature [94,95]. We adhered to established ENVI-met protocols for model inputs and parameter settings. This implementation mirrors successful methodologies in the literature and is fully reproducible

2.5.1. Sample-Area Setup and ENVI-Met Modeling Workflow

Based on the LCZ classification results described above, six sample sites with concentrated suitable green roof areas and strong representativeness were selected from each of LCZ 1, LCZ 2, and LCZ 8 as simulation targets (Figure 4, Figure 5 and Figure 6). Spatial models for each site were constructed using remote sensing imagery, land-use data, and 3D building information, and were then imported into ENVI-met for microclimate simulation.

To ensure standardization and comparability of simulation parameters, surface and building material properties were assigned using ENVI-met’s official material database (DBManager) and applied to the spatial models via the INX plugin.

2.5.2. Key Plant Parameter Values Used in Modeling Different Green Roof Systems

The three green roof strategies defined in this study—extensive, semi-intensive, and intensive—differed significantly in plant-related parameters, including plant height, canopy structure, leaf angle distribution (LAD), and leaf area index (LAI). These differences reflected varying levels of transpiration capacity and shading effects among the roof types [96]. Parameter settings were primarily derived from the simulation reference values proposed by Silva et al. [69], and the corresponding vegetation parameters for the three green roof systems were configured accordingly in DBManager (

Table 3. Vegetation parameter settings in DBManager for the three green roof system.

Green Roof Types	LAI	LAD	Plant Thickness (m)	Substrate Thickness (m)
Extensive	1	0.5	0.1	0.3
Semi-intensive	2.5	0.5	0.5	0.5
Intensive	5	0.5	1	0.7

).

2.5.3. Grid and Meteorological Input Parameter Configuration

ENVI-met simulations employed a three-dimensional cubic grid to define the simulation domain. For each case, the vertical dimension (Z-axis) was set to at least twice the height of the tallest building in the sample area to ensure accurate representation of thermal convection and turbulent diffusion. A five-grid buffer zone was added around the perimeter of each site to stabilize boundary conditions and mitigate edge effects.

Based on these principles, grid configurations were customized for all 18 simulation sites in accordance with local urban morphology.

Meteorological boundary conditions were configured based on a representative extreme summer heat day—20 August 2023—with a 10 h simulation period from 07:00 to 17:00. Initial weather data were obtained from the Dalian TMY file provided by epwmap [97] and imported via the Full Forcing module. Input parameters included dry bulb temperature, humidity, wind speed, wind direction, and solar radiation. Given ENVI-met’s sensitivity to wind field fluctuations, dominant wind directions were adjusted using records from Dalian’s local meteorological station [98] to ensure alignment between simulation outputs and actual climate conditions.

2.5.4. Framework for Analyzing Cooling Effect Differences

For each suitable building within the selected sample sites, four simulation scenarios were defined: (1) original roof (no green cover), (2) extensive green roof, (3) semi-intensive green roof, and (4) intensive green roof. For each scenario, temperature data were extracted from all grid cells, and the average temperature of each sample site was computed.

Within each LCZ category, a Repeated Measures ANOVA was conducted to assess whether statistically significant differences existed among the four roof scenarios. Where significant main effects were detected, paired t-tests were performed to compare roof types, with Bonferroni correction applied to adjust for multiple comparisons.

3. Results

3.1. Spatial Screening and Statistics of Rooftop Suitability for Green Roofs

Based on six key restrictive criteria, this study systematically identified buildings suitable for green roof implementation in Zhongshan and Xigang Districts of Dalian. The screening process proceeded as follows: First, 1102 industrial buildings—primarily constructed with lightweight steel structures and lacking sufficient load-bearing capacity—were excluded. Second, roof accessibility and slope features were automatically extracted using the YOLOv5 object detection model. In this study, we evaluate the object detection performance for three categories of rooftops: sloped rooftops, accessible rooftops, and inaccessible rooftops. The overall detection performance achieves a mean average precision at IoU threshold 0.5 (mAP@0.5) of 0.964. Specifically, the mAP@0.5 for sloped rooftops is 0.954, for accessible flat rooftops is 0.970, and for inaccessible flat rooftops is 0.968. After manual verification, 6803 accessible roofs and 20,618 flat roofs meeting slope criteria were identified. Third, based on historical building records, 589 historically protected buildings were excluded due to legal and preservation constraints. Fourth, rooftop sunlight exposure on the summer solstice was simulated using Ladybug Tools, resulting in the exclusion of 89 buildings that received less than 4 h of average daily sunlight. Lastly, roof surface wind speed was simulated using the Phoenics software. No buildings in the study area exhibited extreme wind speeds exceeding 20 m/s; therefore, no exclusions were made based on wind speed. An overview of the screening criteria and results is presented in Figure 7.

After the model inference was completed, we conducted manual validation to ensure classification accuracy. Using high-resolution Google Earth imagery, a random sample of 1000 detected rooftops—covering all three categories (accessible flat roofs, inaccessible flat roofs, and pitched roofs)—was visually inspected. Misclassifications, primarily caused by partial obstructions from rooftop equipment or shadows, accounted for approximately 13.2% of the sample. These errors were corrected by reassigning the corresponding category labels in the detection results prior to the subsequent suitability analysis. As a result of the multi-criteria screening applied to 28,713 buildings in the study area, a total of 6610 buildings were identified as meeting the suitability requirements for green roof development. The total suitable rooftop area was approximately 2,441,340 m², accounting for 34.4% of the total building area in the region. The spatial distribution and area statistics of suitable buildings are presented in Figure 8.

3.2. Evaluation of Ecological and Economic Benefits of Urban Green Roofs

Building upon the screening results, this study applied the monetization framework established in Section 2.3 to calculate the comprehensive ecological value of the three types of green roofs and conducted a life-cycle cost–benefit analysis. The analysis estimated the NPV and BCR for each roof type under two scenarios: (1) an ideal scenario, in which all ecological benefits were monetized, and (2) a realistic scenario, in which only the direct economic returns from energy conservation and emission reduction were included.

The results indicated that the total annual ecological benefits of green roofs across all suitable rooftops in the study area were as follows: extensive roofs—39,815,210 USD, semi-intensive roofs—66,579,102 USD, and intensive roofs—73,426,702 USD (see

Table 4. Estimated total ecological benefits of three types of green roofs within the suitable construction range.

	Extensive ($/y)	Semi-Intensive ($/y)	Intensive ($/y)
Carbon Sequestration Value	93,688	212,958	304,804
Oxygen Release Value	218,407	724,409	975,374
Air Purification Value	34,462,337	51,723,215	55,110,031
Rainwater Retention Value	290,776	484,626	484,626
Building Energy Conservation	4,750,000	13,433,956	16,551,865
Total	39,815,210	66,579,102	73,426,702

As of 17 August 2025, the exchange rate was 1 USD = 7.1986 CNY.

). Among these, air purification contributed the most to total value, while energy conservation also accounted for substantial proportions.

Under the ideal scenario, all three green roof strategies demonstrated good economic feasibility. The intensive green roof yielded the highest NPV, while the semi-intensive green roof achieved the highest BCR, indicating the most cost-efficient performance. All three types were projected to achieve payback within 4 years (

Table 5. NPV and BCR results for the three green roof types under the ideal scenario.

	Extensive	Semi-Intensive	Intensive
Economic Value	39,815,210 $/y	66,579,102 $/y	528,566,526 $/y
Construction Cost	118,699,990 $	171,945,414 $	217,390,553 $
Maintenance Cost	4,917,571 $	7,121,999 $	7,121,999 $
Project Lifetime	40	40	40
Discount Rate	3%	3%	3%
NPV	687,950,980 $	1,202,391,969 $	1,315,227,547 $
BCR	3.96	4.57	4.44
Payback Period	4	3	4

).

Under the realistic scenario, the results (

Table 6. NPV and BCR results for the three green roof types under the realisticl scenario.

	Extensive	Semi-Intensive	Intensive
Economic Value	4,750,000 $/y	13,433,956 $/y	16,551,865 $/y
Construction Cost	118,699,990 $	171,945,414 $	217,390,553 $
Maintenance Cost	4,917,571 $	7,121,999 $	7,121,999 $
Project Lifetime	40	40	40
Discount Rate	3%	3%	3%
NPV	−122,574,098 $	−26,049,568 $	579,429 $
BCR	0.472	0.922	1.002
Payback Period	Unprofitable	Unprofitable	23

) showed that relying solely on energy-saving benefits, extensive and semi-intensive green roofs were unlikely to achieve payback within their life cycle. While the intensive green roof could reach break-even, the payback period extended to 23 years. These findings indicated that, in the absence of ecological compensation or policy support, the market-based return on investment for green roofs faced significant limitations.

Notably, prices in both the ideal and realistic scenarios are not fixed; they evolve with market conditions. Accordingly, we conducted sensitivity analyses on key parameters—discount rate, energy prices, and ecological value coefficients. The results indicate that, under the ideal scenario, all three green-roof types remain profitable across plausible price variations and their BCR ranking is unchanged (

Table A4. Sensitivity analyses on key parameters under the ideal scenario.

Parameters Changes	Green Roof Type	NPV	BCR	Payback Period
An overall Increase of 10% in all types of ecological benefits prices except energy prices	Extensive	769,003,360 $	4.31	4
	Semi-intensive	1,325,237,334 $	4.94	3
	Intensive	1,446,692,376 $	4.79	4
An overall decrease of 10% in all types of ecological benefits prices except energy prices	Extensive	606,003,360 $	3.61	4
	Semi-intensive	1,079,549,461 $	4.21	4
	Intensive	1,183,762,607 $	4.1	4
10% Increase in both construction and maintenance costs	Extensive	664,714,126 $	3.6	4
	Semi-intensive	1,168,735,084 $	4.16	4
	Intensive	1,277,026,144 $	4.04	4
10% Decrease in both construction and maintenance costs	Extensive	711,187,830 $	4.4	4
	Semi-intensive	1,236,048,844 $	5.08	3
	Intensive	1,353,428,931 $	4.94	3
Decrease in discount rate by 1%	Extensive	835,941,649 $	4.3	4
	Semi-intensive	1,454,532,133 $	4.97	3
	Intensive	1,596,406,373 $	4.87	4
Increase in discount rate by 1%	Extensive	572,021,088 $	3.65	4
	Semi-intensive	1,004,875,581 $	4.21	3
	Intensive	1,094,963,441 $	4.06	4

).

By contrast, under the realistic scenario, NPV, BCR, and payback period are highly sensitive to price shifts. Lower construction and maintenance costs, higher market energy prices, and a reduced discount rate render the semi-intensive green roof profitable over its life cycle (

Table A5. Sensitivity analyses on key parameters under the realistic scenario.

Parameters Changes	Green Roof Type	NPV	BCR	Payback Period
10% Increase in Building Energy Conservation costs	Extensive	−111,593,839 $	4	Unprofitable
	Semi-intensive	5,006,316 $	1.015	22
	Intensive	38,837,908 $	1.102	20
10% Decrease in Building Energy Conservation costs	Extensive	−133,552,872 $	0.425	Unprofitable
	Semi-intensive	−57,098,250 $	0.83	Unprofitable
	Intensive	−37,680,609 $	0.9	Unprofitable
10% Increase in both construction and maintenance costs	Extensive	−145,810,208 $	0.43	Unprofitable
	Semi-intensive	−59,702,847 $	0.839	Unprofitable
	Intensive	−37,622,744 $	0.91	Unprofitable
10% Decrease in both construction and maintenance costs	Extensive	−99,336,503 $	0.525	Unprofitable
	Semi-intensive	7,610,912 $	1.025	23
	Intensive	38,780,043 $	1.113	20
Increase in discount rate by 1%	Extensive	−122,016,685 $	0.435	Unprofitable
	Semi-intensive	−47,014,276 $	0.85	Unprofitable
	Intensive	−30,747,348 $	0.914	Unprofitable
Decrease in discount rate by 1%	Extensive	−123,283,975 $	0.513	Unprofitable
	Semi-intensive	721,195 $	1.002	27
	Intensive	40,567,951 $	1.098	23

) and further improve the economic performance of the intensive type. The opposite changes—higher costs, lower energy prices, or a higher discount rate—drive all three green-roof options into non-recoverable territory within the life cycle (Table A5).

3.3. Heat Environment Simulation Results Based on LCZ Zoning

In this study, six representative blocks were selected from each of the three LCZ types—LCZ 1, LCZ 2, and LCZ 8—as sample areas. For each suitable building, four simulation scenarios were defined: the original roof and the three types of green roofs. Using ENVI-met, pedestrian-level temperatures were simulated for each scenario, and the average temperature of each sample area was used as the indicator for inter-group statistical comparison. The simulation results are presented in Figure 9, Figure 10 and Figure 11. It is worth noting that some of the images in this figure contain abrupt green colour blocks such as LCZ1 (5), which represent green areas.This is due to the fact that the output of ENVI-met may be directly assigned default category colours for different object types (surface, buildings, vegetation, etc.) instead of the simulated numerical colours in the LEONARDO visualisation, which, although this is a typical bug for temperature visualisations, does not affect the results and accuracy of the surface temperature data.

Repeated Measures ANOVA revealed that the green roof type had a statistically significant effect on the cooling intensity (p < 0.01). Post hoc paired comparisons (Figure 12) showed that, for LCZ 1 (compact high-rise zone), all three green roof types significantly reduced pedestrian-level temperature compared to the original roof (Bonferroni-adjusted p < 0.05), although the cooling magnitude was the smallest among the three LCZ types. The intensive green roof produced the strongest cooling effect and was significantly different from that of the other two types of roofs.

In LCZ 2, pairwise comparisons among the three green roof types were statistically significant, and the cooling magnitude followed a clear ascending trend, representing the most pronounced temperature gradient among the zones.

For LCZ 8, all three green roof types improved the thermal environment, showing the largest overall cooling effect observed among the LCZs. However, the variability in the cooling performance across sample sites was higher, and only the difference between the semi-intensive and intensive roofs was statistically significant.

Overall, the ENVI-met simulation confirmed the effectiveness of green roofs in regulating urban thermal conditions, with the cooling performance strongly influenced by both the LCZ type and green roof planting strategy.

3.4. Recommendations on Applicability and Configuration of Green Roof Systems

By combining insights from economic evaluations and microclimate simulations, the selection of green roof types should be tailored to the local building conditions, policy contexts, and spatial planning objectives. As summarized in

Table 7. Summary of key advantages and typical applications of the three green roof types.

Green Roof Types	Main Advantages	Recommended Applications
Intensive	It generates the highest long-term net present value (NPV) under the ideal scenario. It provides the most significant and consistent cooling performance. It is the only option capable of achieving long-term investment recovery under the realistic scenario.	It should be prioritized when the goal is to maximize cooling performance and long-term economic returns. It is most suitable for mid- to low-rise commercial zones—such as LCZ 2 and LCZ 8—where structural load capacity is sufficient and policy incentives are more accessible.
Semi-intensive	It achieves the highest benefit–cost ratio (BCR) under the ideal scenario. It requires lower initial investment and structural load capacity compared to the intensive system. It offers a balanced solution by providing notable cooling and ecological benefits at a moderate cost.	It is recommended when economic efficiency is prioritized and policy support is available. It is suitable for buildings—such as typical mid- to high-rise residential structures—where structural or budget constraints limit the use of intensive systems, but performance requirements exceed those of extensive green roofs.
Extensive	It has the lowest construction and maintenance costs. It imposes minimal structural load requirements, making it the most feasible option for retrofitting existing buildings. It enables rapid and large-scale greening implementation.	It is appropriate for projects with limited structural capacity or financial resources. It is ideal for initiatives aiming to rapidly increase urban green coverage over large areas. It is well suited for high-density, high-rise urban zones—such as LCZ 1.

, intensive green roofs are recommended for scenarios that prioritize maximum cooling and long-term economic returns, particularly where structural and policy support is robust. Semi-intensive systems provide cost-effective alternatives when policy incentives are available; however, intensive solutions are limited by structural and financial constraints. Extensive green roofs are best suited for widespread implementation in areas requiring rapid retrofitting and minimal structural intervention. Notably, in highly heterogeneous urban zones such as LCZ 8, the thermal performance of green roofs may vary substantially among buildings, underscoring the importance of adaptive, site-specific design and deployment. Combining the results of the economic analysis and microclimate simulations, each of the three green roof types demonstrates distinct advantages and is suited to different policy environments, building structures, and urban spatial contexts. The intensive green roof exhibited the best performance in terms of long-term economic returns and thermal regulation capacity. It is particularly suitable for mid-rise or low-rise commercial zones with sufficient load-bearing capacity and strong policy incentives, such as LCZ 2 and LCZ 8. The semi-intensive green roof offers a favorable cost–performance ratio in scenarios where policy support is moderate or intensive roofs are structurally constrained, making it well suited for areas where economic efficiency is prioritized and the structural capacity is average. Extensive green roofs stand out for their low cost and ease of retrofitting, making them ideal for high-density high-rise areas with limited load capacity or tight budgets (e.g., LCZ 1), as well as large-scale greening projects requiring rapid implementation. It is important to note that in areas with high spatial heterogeneity—especially LCZ 8—the cooling performance of green roofs varies significantly between buildings. Therefore, the actual implementation should be flexibly adapted to the building and environmental characteristics.

4. Discussion

4.1. Summary of Research Findings with Multi-Dimensional Evaluation

This study, focused on the core urban districts of Dalian, integrated roof suitability screening, quantitative assessment of ecological and economic benefits, life-cycle cost–benefit scenario analysis, and LCZ-based microclimate simulations to systematically examine the spatial configuration, benefit mechanisms, and optimization strategies of urban green roofs. The findings offer quantitative evidence and methodological innovation to support the advancement of green roof practices in coastal cities of northern China.

The multi-dimensional screening framework developed in this study employed six restrictive indicators—building structure, construction year (building age), roof accessibility, roof slope, roof sunlight exposure, and roof surface wind speed—to evaluate rooftop suitability. The application of deep learning-based object detection significantly enhanced the efficiency and accuracy of suitability identification, overcoming the subjectivity and limitations associated with traditional manual interpretation. Compared to prior studies that relied on a limited number of criteria [12,99], the proposed framework represents a clear improvement in both scientific rigor and automation.

The economic evaluation revealed that, under an ideal scenario with robust policy incentives and mechanisms to monetize ecological services, all three green roof types yielded high NPV and BCR, with intensive and semi-intensive roofs demonstrating particularly favorable investment performance. These results align with previous findings from cities such as Seoul and Hong Kong [100,101]. However, the variation in BCR values across studies also reflects differences in contextual factors, including valuation methods, monetization parameters, and assumptions regarding energy prices. Under the realistic scenario, where only energy-saving benefits were considered, most green roof projects did not yield a positive return, emphasizing the need for ecological compensation and fiscal incentives [102], as well as the sensitivity of results to electricity price fluctuations and climate zone differences [103].

Microclimate simulation results showed that all three green roof types effectively reduced pedestrian-level temperatures across LCZs, with intensive roofs delivering the most significant and consistent cooling, particularly in LCZ 2 (compact mid-rise zones). The largest overall cooling magnitude occurred in LCZ 8 (large low-rise zones), although greater variability was observed among sample sites in this category. These findings highlight the crucial role of urban morphology in shaping cooling performance. Our results corroborate previous studies by Luo et al. [104] and Aboelata et al. [105], confirming that cooling capacity is closely tied to vegetation density and substrate depth [106]. Therefore, future urban green roof planning should incorporate LCZ characteristics and tailor greening strategies to the urban structural context.

4.2. Contributions to Methodology and Practice

This study contributes both methodologically and practically to the advancement of green roof planning and implementation.

Methodologically, it developed a transferable urban green roof suitability assessment framework that provides both theoretical support and empirical evidence for the promotion of green roofs in other cities or regions. For instance, in terms of climate, this framework can be applied to cities with varying temperature, humidity, and precipitation patterns. The influence of certain rooftop suitability criteria, such as sunlight exposure, may be amplified by the climate conditions of different regions. For example, in cities like Chongqing and Chengdu, which experience relatively low sunlight throughout the year, the rooftop sunlight exposure criterion may become particularly critical in the suitability assessment process. Regarding urban morphology, the framework can be adapted to different urban layouts. In high-density cities like Hong Kong, the LCZ classification for rooftop suitability may primarily focus on LCZ1 and exclude other zones such as LCZ2. Conversely, in cities like Dongguan and Zhongshan, where a significant proportion of the urban area consists of low-rise, densely packed urban villages, including LCZ3 in the study framework becomes essential to account for the unique characteristics of these areas. Lastly, the policy environment plays a critical role in the adoption and success of green roofs. The framework’s dual-scenario cost–benefit analysis is particularly useful for cities with varying levels of policy support. In cities with robust ecological compensation mechanisms and incentive policies, semi-intensive green roofs are more likely to generate favorable economic returns. However, in cities where subsidies are limited or policies are weaker, strategies focusing on lower construction and maintenance costs, such as extensive green roofs, may be more suitable.

Practically, the findings offer actionable tools for governments, developers, and urban planners. The suitability assessment framework aids in identifying buildings with the highest retrofitting potential and prioritizing intervention areas. The life-cycle cost–benefit analysis delivers clear investment benchmarks under different scenarios for policymakers, building owners, and investors, thereby supporting informed decision-making and efficient resource allocation—particularly in designing subsidy policies and incentive mechanisms. The LCZ-based microclimate simulation provides zoning-specific configuration guidance for urban planning and heat mitigation, recommending the targeted deployment of intensive green roofs in zones with the highest cooling potential to maximize ecological value. To illustrate how the framework informs decisions under real-world constraints, we offer two hypothetical scenarios. First, in areas dominated by large low-rise commercial buildings (LCZ 8) with strong structural capacity, where local governments or firms operate without a mature ecological-subsidy scheme yet seek substantial cooling alongside partial cost recovery, the framework recommends intensive green roofs: they deliver the largest and more stable cooling effects and can approach—or achieve—financial feasibility when cost and energy-price conditions are favorable. Second, in older residential districts (LCZ 2) with limited load-bearing capacity, where low cost, low maintenance, rapid roll-out, and minimal disruption are priorities while still providing neighborhood-scale cooling, the framework supports extensive green roofs as the preferred option, offering a balanced cost–benefit profile under tight budgetary and construction constraints. Together, these examples show that by aligning structural conditions and policy context with cooling needs and economic targets, model outputs can be translated into actionable, context-specific deployment strategies.

4.3. Specific Policies That Are Financially Viable for “Realistic” Scenarios

In the real world, it can be tough to get societies to have well-developed policies for subsidizing the ecological value of green infrastructure, and this can be even tougher for developing countries such as China. Therefore, to improve financial viability under market-only (“realistic”) conditions, Local governments can learn from existing policies in other countries or regions in two main ways. First, capital subsidies paid per unit roof area lower capex and have been shown to stimulate uptake (e.g., Toronto’s eco-roof grants of $100/m² plus structural-assessment support [107]). Second, tax expenditures—such as property-tax abatements or accelerated depreciation—convert future savings into near-term value (e.g., New York City’s one-year abatement of $10/ft² of green roof [108])

4.4. Limitations and Future Prospects

Despite integrating multi-source data and multiple analytical models, this study has some limitations.

First, the suitability assessment relied primarily on remote sensing and AI-based automated interpretation, without on-site verification of structural attributes such as actual load-bearing capacity. Because most buildings and residential compounds in Dalian enforce strict access controls and restrict rooftop entry, in situ rooftop validation was not feasible. Consequently, some rooftop suitability classifications may be subject to misjudgment. Future research should incorporate drone-based surveys, construction documentation, and other data sources to enhance evaluation accuracy.

Second, although ENVI-met provides high-resolution microclimate simulations, it has limitations in capturing the ecological complexity of vegetation. Specifically, ENVI-met currently allows only a single set of plant parameters, excluding convective heat transfer between leaves and air or radiative heat exchange between plant surfaces and their surroundings [109]. This simplification may lead to biases in simulating physiological effects, particularly under low light or high solar radiation conditions at midday [92]. Additionally, the software permits only a single wall material for all façades and rooftops, or requires approximate thermal substitutions, which oversimplifies urban heterogeneity [110]. At the same time, ENVI-Met is more suitable for neighborhood-scale thermal environment simulations and lacks large-scale thermal environment simulations at the urban scale. Future improvements should involve multi-model integration. For instance, in the large-scale urban studies, such as those conducted in the Xigang and Shahekou districts of Dalian, high-resolution urban heat environment simulation software like WRF can be employed [111]. Through domain-wide simulations, WRF offers a more comprehensive perspective on the temperature distribution, wind patterns, and energy fluxes across different urban areas following green roof retrofitting. This approach helps to overcome the limitations of localized ENVI-met simulations, providing a more accurate assessment of the urban heat island effect in response to various green roof scenarios.

Third, the ecological benefit parameters and monetization estimates in this study were primarily based on literature and data from other cities, without full adaptation to Dalian’s local climate, species composition, and economic context. In reality, the monetary value of ecosystem services depends on multiple factors, including regional economic levels, public willingness to pay, policy incentives, and the local balance between ecological supply and demand [112,113]. Although this study distinguished between ideal and realistic scenarios, ecological valuations were still derived from empirical averages and substituted parameters. Future work should incorporate long-term field monitoring and localized ecological-economic accounting to refine benefit parameters in accordance with local conditions.

5. Conclusions

Using Dalian’s Zhongshan and Xigang districts as a case study, this paper develops an integrated evaluation framework that couples rooftop suitability screening, ecosystem and economic valuation, dual-scenario life-cycle cost–benefit analysis (CBA), and LCZ-based ENVI-met microclimate simulations. The framework systematically compares the applicability and performance of three green-roof types—extensive, semi-intensive, and intensive—in high-density urban settings. Multi-criteria screening identified 6610 eligible buildings with approximately 2,441,340 m² of usable roof area (34.4% of the study-area building stock), providing a solid spatial basis for city-scale deployment. Under the ideal scenario (full monetization of ecosystem services with sufficient policy support), all three roof types exhibit sound economic feasibility, with 3–4-year payback periods. Under the realistic scenario (accounting only for energy-saving benefits), only the intensive type reaches break-even, with a payback of ~23 years (BCR = 1.002), whereas extensive and semi-intensive systems do not recover costs (BCR = 0.472/0.922).

Microclimate simulations show that cooling performance is jointly conditioned by LCZ and roof type: repeated-measures ANOVA indicates a significant type effect (p < 0.001). Intensive roofs deliver the strongest pedestrian-level cooling, followed by semi-intensive roofs, while extensive roofs achieve the least pronounced reductions. LCZ2 presents a clear, type-wise cooling gradient, whereas LCZ8 yields the largest overall cooling magnitude. Overall, the integrated evaluation and configuration framework proposed in this study establishes a coherent, end-to-end pathway for planning green roofs at the city scale. Methodologically, it links four components that are rarely combined within a single workflow—rooftop suitability detection, ecosystem and economic valuation, dual-scenario life-cycle cost–benefit analysis, and LCZ-based ENVI-met simulations—so that physical feasibility, microclimatic performance, and financial viability can be assessed in a single, consistent system. Substantively, the framework is type-resolved (extensive, semi-intensive, and intensive) and morphology-aware (LCZ), enabling practitioners to translate model outputs into deployable configurations rather than generic recommendations. The result is a decision-ready evidence base that clarifies where each roof type is most effective, what cooling and co-benefits to expect, and under which policy conditions investments are likely to be financially justified.

In addition, the study contributes a reproducible workflow with transparent inputs, parameterization choices, and scenario assumptions, which facilitates auditing, transfer to other cities, and iterative improvement as new data become available. By explicitly contrasting ideal (full monetization and incentives) and realistic (market-only) scenarios, the framework also reveals policy levers—such as ecological compensation or targeted subsidies—that shift projects from marginal to viable. Finally, the LCZ-informed microclimate results provide a physically grounded rationale for spatial targeting and for tailoring roof types to structural and budget constraints.

Future work will extend this contribution by incorporating localized observations (e.g., on-roof temperature, wind, and humidity), higher-resolution and multi-source spatial datasets (e.g., LiDAR, thermal imagery, and building information), and formal uncertainty and sensitivity analysis to bound confidence in the results. It will also examine temporal dynamics (seasonality and extreme heat events), broaden co-benefit accounting (carbon, runoff, air quality, and biodiversity), and evaluate distributional and operational factors (maintenance regimes, social equity, and retrofit logistics). Together, these steps will further strengthen the scientific rigor and contextual adaptability of the framework and enhance its usefulness for policy design and large-scale urban implementation.